Operation-Aware Soft Channel Pruning using Differentiable Masks

TL;DR

This paper introduces a differentiable, data-driven channel pruning method that jointly considers batch normalization and ReLU activations to effectively compress neural networks without additional fine-tuning.

Contribution

It presents a novel approach that learns differentiable masks for channels, enabling soft decisions and joint optimization of model parameters and pruning, outperforming existing methods.

Findings

Achieves higher accuracy with similar resource constraints compared to state-of-the-art methods.

Enables pruning without extra fine-tuning procedures.

Effectively considers operation characteristics for better channel selection.

Abstract

We propose a simple but effective data-driven channel pruning algorithm, which compresses deep neural networks in a differentiable way by exploiting the characteristics of operations. The proposed approach makes a joint consideration of batch normalization (BN) and rectified linear unit (ReLU) for channel pruning; it estimates how likely the two successive operations deactivate each feature map and prunes the channels with high probabilities. To this end, we learn differentiable masks for individual channels and make soft decisions throughout the optimization procedure, which facilitates to explore larger search space and train more stable networks. The proposed framework enables us to identify compressed models via a joint learning of model parameters and channel pruning without an extra procedure of fine-tuning. We perform extensive experiments and achieve outstanding performance in terms of the accuracy of output networks given the same amount of resources when compared with the state-of-the-art methods.